The challenge of developing new concepts
“A lot of times, people don’t know what they want until you show it to them.” Steve Jobs.
Deciding on the best development path for a new product can be challenging. Conjoint analysis is a proven technique for testing alternatives such as: new product concepts, prices, marketing communication and distribution channels before they are implemented in practice.
How does conjoint analysis work?
Consumers are shown alternative concepts and asked to choose between them. The alternatives are created based on a computer generated experimental design. The advantage of this is that the computer sometimes creates alternatives you haven’t even considered yet, and more than that, they are created in a special way – such that we learn the maximum from each choice made.
Better yet, the model doesn’t need to test every alternative possible with the consumer. Once sufficient learning has taken place predictions can be made for hundreds or even thousands of possibilities.
- Quantify preferences for products that don’t exist yet.
- Identify the best combinations of features.
- Determine the relative importance of different product attributes without asking consumers directly.
- Run scenarios in a simulator to see if your changes would have a positive or negative impact.
- Run what if scenarios regarding competitor responses.
- Obtain an indication of willingness to pay for features, and the implied monetary value of each feature.
- Plot theoretical revenue and profit curves in response to price changes.
- Understand price elasticity of demand.
An example of a conjoint simulator tool that you receive after the project completes, to complement the research report.This lets you test different scenarios on your own. See how potential changes to features and prices impact your product’s success.
Learn from consumer choices
Survey respondents are shown various ‘choice sets’ (such as in the example below) or ‘profiles’, constructed according to an efficient experimental design. Their answers are used to estimate the model’s mathematical properties, that are then used in the simulator above.
These can be plain text, or images can be included as a feature in their own right.